Device and method for detecting objects from a video signal

Abstract

A technique is proposed for enabling stable detection of an object even when the contrast of an image is lowered overall or partially. A preprocessing section 12 is added to the upstream stage of an object detecting processing section 13 , and provides an obtained corrected image (image obtained by subjecting an input image to contrast enhancement) as an input image for the object detecting processing section 13 . This facilitates the detection of the object even when the contrast of the entire screen is lowered due to fog or rain. Furthermore, the object detecting processing section 13 detects the object by using not only the corrected image from the preprocessing section 12 but also correction characteristic information 23 containing image characteristics regarding the correction (contrast enhancement). This attempts to reduce an influence of noises enhanced by the contrast enhancement.

What is claimed is: 
     
       1. An object detection device which detects an object that has emerged within a monitoring region, on a basis of an input image from a camera photographing the monitoring region, the object detection device comprising:
 a preprocessor that corrects an input image by enhancing a contrast of an input image, and adjusts an enhancing amount in the contrast of the input image, wherein the enhancing amount increases with increasing a learning rate of a background image in an object-detecting processor; 
 wherein the object-detecting processor:
 detects an object that has emerged within a monitoring region by using the input image corrected by the preprocessor and correction characteristic information, the correction characteristic information including an image characteristic derived from the correction in the preprocessor, 
 updates the background image each time the input image is captured, the background image being compared with the input image in detecting the object, and 
 outputs a location information of the detected object to the preprocessor; and 
 
 wherein the preprocessor further:
 generates the corrected input image by enhancing the contrast of a whole or a part of the input image by using a converting equation based on a cumulative frequency for each luminance level of the input image, 
 outputs the correction characteristic information to the object-detecting processor, and 
 performs another correction to a subsequent input image within a region in which the object was detected, according to the location information. 
 
 
     
     
       2. A method performed by an object detection device which detects an object that has emerged within a monitoring region, on a basis of an input image from a camera photographing the monitoring region, the method comprising:
 correcting by a preprocessor an input image by enhancing a contrast of the input image; 
 adjusting by the preprocessor an enhancing amount in the contrast of the input image in a manner that the enhancing amount increases with increasing a learning rate of a background image at an object detecting step; 
 detecting by an object-detecting processor the object that has emerged within the monitoring region by using the input image corrected in the preprocessing step and correction characteristic information, the correction characteristic information including an image characteristic derived from the correcting step; and 
 updating by the object-detecting processor the background image each time the input image is captured, the background image being compared with the input image in detecting the object; 
 generating the correct input image by enhancing the contrast of a whole or a part of the input image in the correcting step by using a converting equation based on a cumulative frequency for each luminance level of the input image; 
 outputting the correction characteristic information to the object detecting step; and 
 performing another correction to a subsequent input image within a region in which the object was detected according to the location information. 
 
     
     
       3. The object detection device of  claim 1 , wherein the correction characteristic information includes at least one of:
 a width of zone in which the cumulative frequency does not exist in a histogram of the second corrected input image, 
 a minimum luminance in the input image, 
 a maximum luminance in the input image, and 
 a maximum level difference between the input image and the second corrected input image, and 
 wherein the object-detecting processor further determines a binary threshold based on the correction characteristic information, the binary threshold is obtained when binarizing a difference between the background image and the second corrected input image. 
 
     
     
       4. The object detection device of  claim 3 , wherein the preprocessor further:
 generates the second corrected input image by calculating a weighted average of the first corrected input image and the input image, and 
 outputs the second corrected input image to the object-detecting processor, a background updating rate being used in calculating the weighted average.